[FLASH-USERS] Flash I/O Benchmark fails

[Islam, Tanzima]

Hi,
 I am a new user of the Flash Benchmark I/O code. I have been trying to generate HDF5 checkpoints from the benchmark application. I can compile successfully with parallel HDF5 library but it fails when I try to run the code.

This is my output from the compilation step:
/usr/local/tools/mvapich2-gnu-1.9/bin/mpif90 -c -DN_DIM=3 -DMAXBLOCKS=100 -DIONMAX=13  get_mfluid_property.F90
/usr/local/tools/mvapich2-gnu-1.9/bin/mpif90 -c -DN_DIM=3 -DMAXBLOCKS=100 -DIONMAX=13  flash_release.f
/usr/local/tools/mvapich2-gnu-1.9/bin/mpif90 -c -DN_DIM=3 -DMAXBLOCKS=100 -DIONMAX=13  flash_benchmark_io.F90
/usr/local/tools/mvapich2-gnu-1.9/bin/mpif90 -c -DN_DIM=3 -DMAXBLOCKS=100 -DIONMAX=13  checkpoint_hdf5_parallel.F90
/usr/local/tools/mvapich2-gnu-1.9/bin/mpif90 -c -DN_DIM=3 -DMAXBLOCKS=100 -DIONMAX=13  plotfile_hdf5_parallel.F90
/usr/local/tools/mvapich2-gnu-1.9/bin/mpicc -c -I /usr/local/tools/hdf5-gnu-parallel-1.8.10/include -DTFLOPS -DN_DIM=3 -DH5_USE_16_API h5_file_interface.c
/usr/local/tools/mvapich2-gnu-1.9/bin/mpicc -c -I /usr/local/tools/hdf5-gnu-parallel-1.8.10/include -DTFLOPS -DN_DIM=3 -DH5_USE_16_API h5_parallel_write.c
/usr/local/tools/mvapich2-gnu-1.9/bin/mpicc -c -I /usr/local/tools/hdf5-gnu-parallel-1.8.10/include -DTFLOPS -DN_DIM=3 -DH5_USE_16_API h5_parallel_write_single.c
/usr/local/tools/mvapich2-gnu-1.9/bin/mpif90  -o  flash_benchmark_io get_mfluid_property.o flash_release.o flash_benchmark_io.o  checkpoint_hdf5_parallel.o plotfile_hdf5_parallel.o h5_file_interface.o h5_parallel_write.o h5_parallel_write_single.o -L /usr/local/tools/hdf5-gnu-parallel-1.8.10/lib -lhdf5 -L/usr/lib -lz

I get the following error message when I try to run the benchmark:

bash-4.1$ srun -N 4 -n 4 ./flash_benchmark_io                   ####### Running the code

Fatal error in PMPI_Bcast: Invalid datatype, error stack:
PMPI_Bcast(1549): MPI_Bcast(buf=0x7fffffffc890, count=8, INVALID DATATYPE, root=0, comm=0x84000001) failed
PMPI_Bcast(1501): Invalid datatype
Fatal error in PMPI_Bcast: Invalid datatype, error stack:
PMPI_Bcast(1549): MPI_Bcast(buf=0x7fffffffc890, count=8, INVALID DATATYPE, root=0, comm=0x84000001) failed
PMPI_Bcast(1501): Invalid datatype
Fatal error in PMPI_Bcast: Invalid datatype, error stack:
PMPI_Bcast(1549): MPI_Bcast(buf=0x7fffffffc890, count=8, INVALID DATATYPE, root=0, comm=0x84000001) failed
PMPI_Bcast(1501): Invalid datatype
Fatal error in PMPI_Bcast: Invalid datatype, error stack:
PMPI_Bcast(1549): MPI_Bcast(buf=0x7fffffffc890, count=8, INVALID DATATYPE, root=0, comm=0x84000001) failed
PMPI_Bcast(1501): Invalid datatype
 HDF 5 v 1.4 test on Blue Pacific
           4  processors
 nxb, nyb, nzb =           16          16          16
 nguard =            4
 number of blocks =           80
 nvar =           24

 doing HDF 5 parallel I/O to a single file
srun: error: sierra27: task 3: Exited with exit code 1

I am not sure if I have to supply anything via compilation flag to specify the datatype. Any idea?

Will appreciate any answer.

Regards,
--Tanzima
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20130909/36310cb8/attachment.htm>